README

Linux Malware Detect v1.5
            (C) 2002-2016, R-fx Networks <proj@r-fx.org>
            (C) 2016, Ryan MacDonald <ryan@r-fx.org>
This program may be freely redistributed under the terms of the GNU GPL v2

::::::::::::::::::::::::::::::::::

:: CONTENTS ::
.: 1 [ DESCRIPTION ]
.: 2 [ FEATURES ]
.: 3 [ THREAT SOURCE DATA ]
.: 4 [ RELEASE UPDATES ]
.: 4.1 [ SIGNATURE UPDATES ]
.: 5 [ DETECTED THREATS ]
.: 6 [ THREAT SHARING ]
.: 7 [ CONFIGURATION ]
.: 8 [ IGNORE OPTIONS ]
.: 9 [ CLI USAGE ]
.: 10 [ CRON DAILY ]
.: 11 [ INOTIFY MONITORING ]
.: 12 [ MODSECURITY2 UPLOAD SCANNING ]
.: 13 [ CLEANER RULES ]

::::::::::::::::::::::::::::::::::

.: 1 [ DESCRIPTION ]

Linux Malware Detect (LMD) is a malware scanner for Linux released under the 
GNU GPLv2 license, that is designed around the threats faced in shared hosted 
environments. It uses threat data from network edge intrusion detection 
systems to extract malware that is actively being used in attacks and 
generates signatures for detection. In addition, threat data is also derived 
from user submissions with the LMD checkout feature and from malware 
community resources. The signatures that LMD uses are MD5 file hashes and HEX 
pattern matches, they are also easily exported to any number of detection 
tools such as ClamAV.

The driving force behind LMD is that there is currently limited availability 
of open source/restriction free tools for Linux systems that focus on malware 
detection and more important that get it right. Many of the AV products that 
perform malware detection on Linux have a very poor track record of detecting 
threats, especially those targeted at shared hosted environments.

The threat landscape in shared hosted environments is unique from that of the 
standard AV products detection suite in that they are detecting primarily OS 
level trojans, rootkits and traditional file-infecting viruses but missing 
the ever increasing variety of malware on the user account level which serves 
as an attack platform.

Using the CYMRU malware hash registry, which provides malware detection data 
for 30 major AV packages, we can demonstrate this short coming in current 
threat detection. The following is an analysis of 8,883 MD5 hashes that ship 
in LMD 1.5 and the percentage of major AV products that currently detect 
the hashes.

KNOWN MALWARE:      1951
 % AV DETECT (AVG):  58
 % AV DETECT (LOW):  10
 % AV DETECT (HIGH): 100
UNKNOWN MALWARE:    6931

What this information means, is that of the 8,883 hashes, 78% or 6,931 malware threats
are NOT detected by top-30 AV products. The 1,951 detected malware threats that are known
have an average detection rate of 58% among top-30 AV products with a low and high
detection rate of 10% and 100% respectively. This clearly demonstrates the significant
lapse in user space malware detection that top-30 AV products currently provide. It is for
this reason LMD was created, to fill a void, specifically for shared hosted environments.

.: 2 [ FEATURES ]

- MD5 file hash detection for quick threat identification
- HEX based pattern matching for identifying threat variants
- statistical analysis component for detection of obfuscated threats (e.g: base64)
- integrated detection of ClamAV to use as scanner engine for improved performance
- integrated signature update feature with -u|--update
- integrated version update feature with -d|--update-ver
- scan-recent option to scan only files that have been added/changed in X days
- scan-all option for full path based scanning
- checkout option to upload suspected malware to rfxn.com for review / hashing
- full reporting system to view current and previous scan results
- quarantine queue that stores threats in a safe fashion with no permissions
- quarantine batching option to quarantine the results of a current or past scans
- quarantine restore option to restore files to original path, owner and perms
- quarantine suspend account option to Cpanel suspend or shell revoke users
- cleaner rules to attempt removal of malware injected strings
- cleaner batching option to attempt cleaning of previous scan reports
- cleaner rules to remove base64 and gzinflate(base64 injected malware
- daily cron based scanning of all changes in last 24h in user homedirs
- daily cron script compatible with stock RH style systems, Cpanel & Ensim
- kernel based inotify real time file scanning of created/modified/moved files
- kernel inotify monitor that can take path data from STDIN or FILE
- kernel inotify monitor convenience feature to monitor system users
- kernel inotify monitor can be restricted to a configurable user html root
- kernel inotify monitor with dynamic sysctl limits for optimal performance
- kernel inotify alerting through daily and/or optional weekly reports
- HTTP upload scanning through mod_security2 inspectFile hook
- e-mail alert reporting after every scan execution (manual & daily)
- path, extension and signature based ignore options
- background scanner option for unattended scan operations
- verbose logging & output of all actions


.: 3 [ THREAT SOURCE DATA ]

The defining difference with LMD is that it doesn't just detect malware based 
on signatures/hashes that someone else generated but rather it is an 
encompassing project that actively tracks in the wild threats and generates 
signatures based on those real world threats that are currently circulating.

There are four main sources for malware data that is used to generate LMD 
signatures:
- Network Edge IPS: Through networks managed as part of my day-to-day job,
primarily web hosting related, our web servers receive a large amount of daily
abuse events, all of which is logged by our network edge IPS. The IPS events
are processed to extract malware url's, decode POST payload and base64/gzip
encoded abuse data and ultimately that malware is retrieved, reviewed, classified
and then signatures generated as appropriate. The vast majority of LMD signatures
have been derived from IPS extracted data.

The network I manage hosts over 35,000 web sites and as 
such receives a large amount of daily abuse, all of which is logged by our 
network edge IPS. The IPS events are processed to extract malware url's, 
decode POST payload and base64/gzip encoded abuse data and ultimately that 
malware is retrieved, reviewed, classified and then signatures generated as 
appropriate. The vast majority of LMD signatures have been derived from IPS 
extracted data.
- Community Data: Data is aggregated from multiple community malware websites 
such as clean-mx and malwaredomainlist then processed to retrieve new 
malware, review, classify and then generate signatures.
- ClamAV: The HEX & MD5 detection signatures from ClamAV are monitored for 
relevant updates that apply to the target user group of LMD and added to the 
project as appropriate. To date there has been roughly 400 signatures ported 
from ClamAV while the LMD project has contributed back to ClamAV by 
submitting over 1,100 signatures and continues to do so on an ongoing basis.
- User Submission: LMD has a checkout feature that allows users to submit 
suspected malware for review, this has grown into a very popular feature and 
generates on average about 30-50 submissions per week.

.: 4 [ RELEASE UPDATES ]
Updates to the release version of LMD are not automatically installed but can
be installed using the --update-ver option. There is good reasons that this is
not done automatically and I really dont feel like listing them so just think
about it a bit.

The latest changes in the release version can always be viewed at:
http://www.rfxn.com/appdocs/CHANGELOG.maldetect

.: 4.1 [ SIGNATURE UPDATES ]

The LMD signatures are updated typically once per day or more frequently
depending on incoming threat data from the LMD checkout feature, IPS malware
extraction and other sources. The updating of signatures in LMD installations
is performed daily through the default cron.daily script with the --update
option, which can be run manually at any time.

An RSS & XML data source is available for tracking malware threat updates:
RSS Recent Signatures: http://www.rfxn.com/api/lmd
XML Recent Signatures: http://www.rfxn.com/api/lmd?id=recent
XML All Signatures:    http://www.rfxn.com/api/lmd?id=all

.: 5 [ DETECTED THREATS ]

LMD 1.5 has a total of 10,796 (8883 MD5 / 1913 HEX) signatures (before updates),
below is a listing of the top 60 threats by prevalence detected by LMD.

base64.inject.unclassed    bin.dccserv.irsexxy      bin.fakeproc.Xnuxer
bin.ircbot.nbot            bin.ircbot.php3          bin.ircbot.unclassed
bin.pktflood.ABC123        bin.pktflood.osf         bin.trojan.linuxsmalli
c.ircbot.tsunami           exp.linux.rstb           exp.linux.unclassed
exp.setuid0.unclassed      gzbase64.inject          html.phishing.auc61
html.phishing.hsbc         perl.connback.DataCha0s  perl.connback.N2
perl.cpanel.cpwrap         perl.mailer.yellsoft     perl.ircbot.atrixteam
perl.ircbot.bRuNo          perl.ircbot.Clx          perl.ircbot.devil
perl.ircbot.fx29           perl.ircbot.magnum       perl.ircbot.oldwolf
perl.ircbot.putr4XtReme    perl.ircbot.rafflesia    perl.ircbot.UberCracker
perl.ircbot.xdh            perl.ircbot.xscan        perl.shell.cbLorD
perl.shell.cgitelnet       php.cmdshell.c100        php.cmdshell.c99
php.cmdshell.cih           php.cmdshell.egyspider   php.cmdshell.fx29
php.cmdshell.ItsmYarD      php.cmdshell.Ketemu      php.cmdshell.N3tshell
php.cmdshell.r57           php.cmdshell.unclassed   php.defash.buno
php.exe.globals            php.include.remote       php.ircbot.InsideTeam
php.ircbot.lolwut          php.ircbot.sniper        php.ircbot.vj_denie
php.mailer.10hack          php.mailer.bombam        php.mailer.PostMan
php.phishing.AliKay        php.phishing.mrbrain     php.phishing.ReZulT
php.pktflood.oey           php.shell.rc99           php.shell.shellcomm

.: 6 [ THREAT SHARING ]

I am a firm believer in not reinventing the wheel, for my own sanity or that
of others. As such all unique threat data is submitted to CYMRU & ClamAV so
that the open source and anti-malware community at large can grow from this
project.

.: 7 [ CONFIGURATION ]

The configuration of LMD is handled through /usr/local/maldetect/conf.maldet
and all options are well commented for ease of configuration.

By default LMD has the auto-qurantine of files disabled, this will mean that
YOU WILL NEED TO ACT on any threats detected or pass the SCANID to the '-q'
option to batch quarantine the results. To change this please set quar_hits=1
in conf.maldet.

.: 8 [ IGNORE OPTIONS ]

There are four ignore files available and they break down as follows:

/usr/local/maldetect/ignore_paths
A line spaced file for paths that are to be execluded from search results
 Sample ignore entry:
 /home/user/public_html/cgi-bin

/usr/local/maldetect/ignore_file_ext
A line spaced file for file extensions to be excluded from search results
 Sample ignore entry:
 .js
 .css

/usr/local/maldetect/ignore_sigs
A line spaced file for signatures that should be removed from file scanning
 Sample ignore entry:
 base64.inject.unclassed

/usr/local/maldetect/ignore_inotify
A line spaced file for regexp paths that are excluded from inotify monitoring
 Sample ignore entry:
 ^/home/user$
 ^/var/tmp/#sql_.*\.MYD$

.: 9 [ CLI USAGE ]

Once LMD is installed it can be run through the 'maldet' command, the '--help'
option gives a detailed summary of usage options:

    -b, --background
      Execute operations in the background, ideal for large scans
      e.g: maldet -b -r /home/?/public_html 7

    -u, --update
       Update malware detection signatures from rfxn.com

    -d, --update-ver
       Update the installed version from rfxn.com

    -m, --monitor USERS|PATHS|FILE
       Run maldet with inotify kernel level file create/modify monitoring
       If USERS is specified, monitor user homedirs for UID's > 500
       If FILE is specified, paths will be extracted from file, line spaced
       If PATHS are specified, must be comma spaced list, NO WILDCARDS!
       e.g: maldet --monitor users
       e.g: maldet --monitor /root/monitor_paths
       e.g: maldet --monitor /home/mike,/home/ashton

    -k, --kill
       Terminate inotify monitoring service

    -r, --scan-recent PATH DAYS
       Scan files created/modified in the last X days (default: 7d, wildcard: ?)
       e.g: maldet -r /home/?/public_html 2

    -a, --scan-all PATH
       Scan all files in path (default: /home, wildcard: ?)
       e.g: maldet -a /home/?/public_html

    -c, --checkout FILE
       Upload suspected malware to rfxn.com for review & hashing into signatures

    -l, --log
       View maldet log file events

    -e, --report SCANID email
       View scan report of most recent scan or of a specific SCANID and optionally
       e-mail the report to a supplied e-mail address
       e.g: maldet --report
       e.g: maldet --report list
       e.g: maldet --report 050910-1534.21135
       e.g: maldet --report SCANID user@domain.com

    -s, --restore FILE|SCANID
       Restore file from quarantine queue to orginal path or restore all items from
       a specific SCANID
       e.g: maldet --restore /usr/local/maldetect/quarantine/config.php.23754
       e.g: maldet --restore 050910-1534.21135

    -q, --quarantine SCANID
       Quarantine all malware from report SCANID
       e.g: maldet --quarantine 050910-1534.21135

    -n, --clean SCANID
       Try to clean & restore malware hits from report SCANID
       e.g: maldet --clean 050910-1534.21135

    -U, --user USER
       Set execution under specified user, ideal for restoring from user quarantine or
       to view user reports.
       e.g: maldet --user nobody --report
       e.g: maldet --user nobody --restore 050910-1534.21135

    -co, --config-option VAR1=VALUE,VAR2=VALUE,VAR3=VALUE
       Set or redefine the value of conf.maldet config options
       e.g: maldet --config-option email_addr=you@domain.com,quar_hits=1

    -p, --purge
       Clear logs, quarantine queue, session and temporary data.

.: 10 [ CRON DAILY ]

The cronjob installed by LMD is located at /etc/cron.daily/maldet and is used
to perform a daily update of signatures, keep the session, temp and quarantine
data to no more than 14d old and run a daily scan of recent file system changes.

The daily scan supports Ensim virtual roots or standard Linux /home*/user paths,
such as Cpanel. The default is to just scan the web roots daily, which breaks
down as /home*/*/public_html or on Ensim /home/virtual/*/fst/var/www/html and
/home/virtual/*/fst/home/*/public_html.

If you are running monitor mode, the daily scans will be skipped and instead a
daily report will be issued for all monitoring events. If you need to scan
additional paths, you should review the cronjob and edit it accordingly.

.: 11 [ INOTIFY MONITORING ]

The inotify monitoring feature is designed to monitor users in real-time for
file creation/modify/move operations. This option requires a kernel that
supports inotify_watch (CONFIG_INOTIFY) which is found in kernels 2.6.13+
and CentOS/RHEL 5 by default. If you are running CentOS 4 you should consider
an inbox upgrade with: http://www.rfxn.com/upgrade-centos-4-8-to-5-3/

There are three modes that the monitor can be executed with and they relate
to what will be monitored, they are USERS|PATHS|FILES. 
       e.g: maldet --monitor users
       e.g: maldet --monitor /root/monitor_paths
       e.g: maldet --monitor /home/mike,/home/ashton

The options break down as follows:
USERS -  The users option will take the homedirs of all system users that are
         above inotify_minuid and monitor them. If inotify_webdir is set then
         the users webdir, if it exists, will only be monitored.
PATHS -  A comma spaced list of paths to monitor
FILE  -  A line spaced file list of paths to monitor

Once you start maldet in monitor mode, it will preprocess the paths based on
the option specified followed by starting the inotify process. The starting of
the inotify process can be a time consuming task as it needs to setup a monitor
hook for every file under the monitored paths. Although the startup process can
impact the load temporarily, once the process has started it maintains all of
its resources inside kernel memory and has a very small userspace footprint in
memory or cpu usage.

The scanner component of the monitor watches for notifications from the inotify
process and batches items to be scanned, by default, every 30 seconds. If you
need tighter control of the scanning timer, you can edit inotify_stime in
conf.maldet.

The alerting of file hits under monitor mode is handled through a daily report
instead of sending an email on every hit. The cron.daily job installed by LMD
will call an --alert-daily flag and send an alert for the last days hits. There
is also an --alert-weekly option that can be used, simply edit the cron at
/etc/cron.daily/maldet and change the --alert-daily to --alert-weekly.

Terminating the inotify monitoring is done by passing the '-k|--kill-monitor'
option to maldet, it will touch a file handle monitored by maldet and on the
next waking cycle of the monitor service, it will terminate itself and all
inotify processes.

.: 12 [ MODSECURITY2 UPLOAD SCANNING ]

The support for HTTP upload scanning is provided through mod_security2's inspectFile hook.
This feature allows for a validation script to be used in permitting or denying an upload. 

The convenience script to faciliate this is called hookscan.sh and is located in the
/usr/local/maldetect installation path. The default setup is to run a standard maldet scan
with no clamav support, no cleaner rule executions and quarantining enabled; these options
are set in the interest of performance vs accuracy which is a fair tradeoff. 

The scan options can be modified in the hookscan.sh file if so desired, the default
scan options are as follows:
--config-option quar_hits=1,quar_clean=0,clamav_scan=0 --modsec -a "$file"

There is a tangible performance difference in disabling clamav scanning in this usage
scenario. The native LMD scanner engine is much faster than the clamav scanner engine
in single file scans by a wide margin. A single file scan using clamav takes roughly
3sec on average while the LMD scanner engine takes 0.5sec or less.

To enable upload scanning with mod_security2 you must set enable the public_scan option
in conf.maldet (public_scan=1) then add the following rules to your mod_security2 
configuration. These rules are best placed in your modsec2.user.conf file on cpanel servers
or at the top of the appropraite rules file for your setup.

/usr/local/apache/conf/modsec2.user.conf (or similar mod_security2 rules file):
SecRequestBodyAccess On
SecRule FILES_TMPNAMES "@inspectFile /usr/local/maldetect/hookscan.sh" \
                "id:'999999',log,auditlog,deny,severity:2,phase:2,t:none"

A restart of the HTTPd service is required following these changes.

When an upload takes place that is determined to be malware, it will be rejected and an
entry will appear in the mod_security2 SecAuditLog file. On cpanel servers and most
configurations this is the modsec_audit.log located under /usr/local/apache/logs or 
/var/log/httpd.

The log entry will appear similar to the following:
Message: Access denied with code 406 (phase 2). File "/tmp/20121120-....-file" rejected by
the approver script "/usr/local/maldetect/hookscan.sh": 0 maldet: {HEX}php.cmdshell.r57.317
/tmp/20121120-....-file [file "/usr/local/apache/conf/modsec2.user.conf"] [line "3"]
[severity "CRITICAL"]

The default alerting options will apply and an e-mail will be sent when hits are found. This
can be changed in the hookscan.sh script by editing the --config-option values.

To disable alerts append email_alert=0 to the --config-option values:
--config-option quar_hits=1,quar_clean=0,clamav_scan=0,email_alert=0

To change the e-mail address for alerts on upload hits, append email_addr=you@domain.com
to the --config-option values:
--config-option quar_hits=1,quar_clean=0,clamav_scan=0,email_addr=you@domain.com

The nature of uploads is such that they are performed either under the user that the HTTP
service is running as or under that of a system user in an suexec style setup (i.e: phpsuexec).
This required a change to the way LMD stores session, temporary and quarantine data to allow
for non-root users to perform scans.

Given that the maldetect installation path is owned by user root, we either need to set a pub
path world writable (777) or populate the pub path with user owned paths. It was undesirable
to set any path world writable and as such a feature to populate path data was created. This
feature is controlled with the --mkpubpaths flag and is executed from cron every 10 minutes,
it will only execute if the public_scan variable is enabled in conf.maldet. As such, it is
important to make sure the public_scan variable is set to enabled (1) in conf.maldet and it is
advised to run 'maldet --mkpubpaths' manually to prepopulate the user paths. There after, the
cron will ensure new users have paths created no later than 10 minutes after creation.

All non-root scans, such as those performed under mod_security2, will be stored under the
/usr/local/maldetect/pub/username directory tree. The quarantine paths are relative to the user
that exectues the scan, so user nobody would be under pub/nobody/quar/. The actual paths
for where files are quarantined and the user which executed the scan, can be verified in the
e-mail reports for upload hits.

To restore files quarantined under non-root users, you must pass the -U|--user option to LMD,
for example if user nobody quarantined a file you would like to restore, it can be restored as
follows:
maldet --user nobody /usr/local/maldetect/pub/nobody/quar/20121120-file-SFwTeu.22408

Or, as always the scan ID can be used to restore
maldet --user nobody 112012-0032.13771

.: 13 [ CLEANER RULES ]

The cleaner function looks for signature-named rules under the clean/ path,
these rules can consist of any command that is designed to clean a file of
malware. A cleaner rule must result in a file being able to pass a scan
without tripping a HIT otherwise it will classify the clean action as FAILED.

Let us assume for a moment we have malware that we want to clean and it trips
with the signature "{HEX}php.cmdshell.r57.89". The actual signature string in
this is "php.cmdshell.r57", the "{HEX}" just defines the format and ".89" is
the variant number. So, to create a clean rule for php.cmdshell.r57 we would
add a file 'clean/php.cmdshell.r57' and this would be executed against any
file that hits on the signature of the same name.

The actual contents of the rule should be a single line command that will be
executed against the hit file, for example the execution looks something like:

YOUR_COMMAND MALWARE_FILE

So, for a string based malware injection you could easily throw in a 'sed -i'
into the rule file with the appropriate pattern to strip the string(s) from
the file. Once the clean command has run, a rescan will be performed on the
file and if it causes causes a hit, the clean will be marked as FAILED. A
successful clean ALWAYS results in the file being restored if possible to
its original path, owner and mode.

An important note is that the cleaner function is a subfunction of the
quarantine, so if the quarantine is disabled then by default, malware hits
will not have clean attempts made. There are two ways around this, apart from
the obvious of turning on quarantine and rescanning (which is a waste of time).
The best way is to enable the quarantine and then use the -q|--quarantine flag
to batch through the scan results, which will quarantine and clean files. The
second is to use the -n|--clean flag which will try to clean files in place,
be that in the quarantine or the files original path, wherever it can be found.

e.g: maldet -q SCANID
e.g: maldet --clean SCANID